A search and discovery stack your team can operate

Choose a search and discovery stack by the decisions your team must make and the outcomes it must verify. Map product data, search, merchandising, measurement and support ownership before adding or replacing a tool. Start with a scope that can be delivered, measured and supported within your operating model.

Begin with an observable customer problem

Name the journey that needs improvement: finding an exact item, discovering a suitable alternative, narrowing a large category or completing a repeat purchase. Define the outcome and the evidence that would show progress.

Separate what the search engine can influence from price, assortment, delivery and interface constraints. If a product is unavailable or incorrectly described, changing the ranking model alone may not resolve the customer problem.

Map the owners behind the experience

Create a practical responsibility map for catalogue quality, stock and price updates, retrieval, merchandising rules, front-end behaviour, events and reporting. Identify who approves changes, who resolves incidents and who can restore a prior configuration.

Include commercial stakeholders in that map. Promotions and sponsored placements need clear priorities and constraints so separate teams do not unknowingly change the same customer experience.

Choose the smallest useful change

Consider improving the current configuration, adding a scoped capability or replacing a broader set of components. Compare these options on the same requirements. A consolidated platform can simplify some interfaces while changing operational responsibilities; several tools can also work well when ownership and data contracts are clear.

PA’s Universal Search describes a broader products/content/service scope, while Merchandiser Control covers influence across discovery surfaces. Use those product areas to frame a demonstration and confirm the exact boundaries of the implementation you need.

Build the decision pack

Record the baseline, material shopper tasks, technical constraints, implementation work and recurring responsibilities. Request current evidence for required capabilities and matched commercial proposals. Keep an unanswered requirement visible rather than filling it with a presumed feature or a promised uplift.

Make the exit and expansion paths part of the plan: how catalogue and event data remain available, what can be changed independently and which dependencies a later module introduces. These are questions for the proposal and contract, not conclusions from a product diagram.

Give the first scope an evidence-based review

Evaluate relevance on representative queries, then run the agreed live measurement where appropriate. Include error rates, latency and availability alongside the commercial outcome. Reserve time for unresolved issues and document who accepts them.

Bring the responsibility map and first shopper problem to a PA discussion. The useful next step is a scoped test and an accountable operating plan that your digital, merchandising and engineering teams can support together.

Frequently asked questions

Do we have to replace the whole discovery stack?

Define the required change first. Assess a targeted improvement and a broader replacement against the same shopper tasks, dependencies and costs. Confirm the modular boundaries of the proposed implementation with the supplier.

Who should own search performance?

Name an accountable business owner and explicit technical, data and measurement owners. The exact organisation varies, but important decisions and incidents should not fall between teams.

What belongs in the business case?

Include measured baseline, stated outcome assumptions, implementation and recurring costs, dependencies, risks and the validation plan. Keep an assumed uplift separate from an observed causal result.